Lines of Code
[root]/Packages/Uintah/CCA/Ports
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| Totals | 24 (100.0%) | 4216 (100.0%) | 175.6 |
| jas | 5 (20.8%) | 3695 (87.6%) | 739.0 |
| luitjens | 13 (54.2%) | 494 (11.7%) | 38.0 |
| harman | 2 (8.3%) | 16 (0.4%) | 8.0 |
| dav | 4 (16.7%) | 11 (0.3%) | 2.7 |
Seperated the processes of regridding and load balancing in the simulation controller. This should make it easier to load balance without regridding. Currently this is still not supported due to a few issues when loadbalancing when not also regridding.
3 lines of code changed in 2 files:
Removed the blocked merging operation. The SFC code is now much simpler, easier to understand, easier to debug, and easier to maintain.
The optimizations obscured the code and had little to no effect on performance because they would only be useful if we had 3000+ patches per processor (something that will never happen).
0 lines of code changed in 2 files:
Removed the sample merging. This is part of the simplification process for this code.
8 lines of code changed in 1 file:
Simplifying the merge-exchange operation.
459 lines of code changed in 1 file:
Tasks that require data from a fine level are now profiled sperately from normal tasks. This will hopefully allow the profiling to be more accurate.
2 lines of code changed in 2 files:
Pass the task into the profiler instead of the patches that the task ran on. This will be used to profile the coarse interfaces seperately.
2 lines of code changed in 1 file:
Added a SCI_DEBUG called MPIStats. This debug will printout the number of messages that the task graph sends and the total message volume.
2 lines of code changed in 2 files:
changed label names so they are consist with ICE label
mass_source_CC -> modelMass_src
momentum_source_CC -> modelMom_src
energy_source_CC -> modelEng_src
sp_vol_source_CC -> modelVol_src
density_CC -> rho_CC
velocity_CC -> vel_CC
temperature_CC -> temp_CC
pressure_CC -> press_CC
This should not change the answers
16 lines of code changed in 2 files:
Fixes to allow Uintah to build using the PGI compiler:
M configure
M configure.ac
- Figure out size of 'long long' so that Endian.cc swapbytes() can be
set correctly... perhaps a bit of overkill, but...
- PGI compilers don't support -Wall, so don't use it.
- Better messages when determining dependency generation flag. Fix
grep for ":" as it sometimes is " :". See below (configVars.mk.in).
- pgf77 needs -lrt to link.
M configVars.mk.in
- ECHO should not be "echo -n" (no newline)... so named it more appropriately.
- Many compilers don't create dependency files (.d) with exactly "name.o : name.cc".
Then may use "name.o:name.cc". So don't use the ":" in the sed. This should
never cause a problem because of the "^".
- Added the dependency mode of: 'modify_and_move'. PGI compiler creates an invalid
.d file in the wrong place, so we have to fix it and move it.
- NOTE: the F77 dependency flag is currently (historically) just a copy of the CXX
flag... this has worked in the past and still works for everything but pgf77.
Soon (but not in this update) I will fix configure to do this correctly.
M StandAlone/regression/sub.mk
M Packages/Uintah/StandAlone/tools/compare_mms/sub.mk
M Packages/Uintah/StandAlone/tools/radiusMaker/sub.mk
- Need BLAS_LIBRARY to link... at least on Ranger using PGI.
- Please indent... it really does make it easier to read.
M Core/Malloc/Allocator.cc
Moved variable into the #if section that it is used in. Avoids compiler warning.
M Core/Util/Endian.cc
A Core/Util/testEndian.h.in
D Core/Util/Endian.h
- pgcc is requiring the swapbytes() for 'long long'... So add it
in. Since 'long long' size can be different on different machines,
use overkill method with configure figuring out size and setting it
for us.
- Moved Endian.h to testEndian.h.in to allow configure to modify it.
- Indented.
M Core/Util/sub.mk
M Packages/Uintah/Core/Grid/Variables/sub.mk
Untabified.
M Core/Util/Socket.cc
Fix compiler warning and possible 'random' error.
M Core/Math/MiscMath.cc
Fix PGI compiler issue.
M Packages/Uintah/CCA/Components/DataArchiver/DataArchiver.cc
M Packages/Uintah/CCA/Components/Schedulers/SchedulerCommon.cc
#include <time.h> is required for time functions.
M Packages/Uintah/CCA/Components/ICE/EOS/IdealGas.cc
M Packages/Uintah/CCA/Components/ICE/CustomBCs/sine.h
M Packages/Uintah/CCA/Components/ICE/CustomBCs/MMS_BCs.cc
M Packages/Uintah/CCA/Components/ICE/CustomBCs/microSlipBCs.cc
M Packages/Uintah/CCA/Components/ICE/BoundaryCond.cc
- Removed a bunch of definitions of "iter". (vector<IntVector>::const_iterator)...
probably a leftover from the iterator update a while back. In some cases,
renamed an interior (for loop) iterator from "iter" to "cIter" so that the
code would be a little more clear as to which iterator was being used
inside the inner for loop. Also, define some iterators in the for loop,
instead of above the loop. (for( vector::iterator iter...))
M Packages/Uintah/CCA/Components/ICE/AMRICE.h
M Packages/Uintah/Core/Grid/Variables/GridVariable.h
M Packages/Uintah/Core/Grid/Variables/GridVariableBase.h
Fix compiler warning about hidden parent class function.
M Packages/Uintah/CCA/Components/Arches/ExtraScalarSolver.h
Fix compiler warning about ineffectual use of a const return type. Eg:
const int doSomething() { return 1; }
doesn't mean anything... it doesn't make sense for the returned int to
be const. Should just be:
int doSomething() { return 1; }
M Packages/Uintah/CCA/Ports/SFC.h
Fix problem with nested comments. Don't use /* ... */ over many lines. Use #if 0.
M Packages/Uintah/StandAlone/tools/uda2nrrd/particles.cc
Comment about invalid code.
M Packages/Uintah/StandAlone/tools/puda/varsummary.cc
PGI compiler needs Min/Max defined.
M Packages/Uintah/StandAlone/tools/extractors/partextract.cc
#include <stdlib.h> for strtoll... Add hack to force (older (verion
7.1-2)) pgCC to allow strtoll.
M Packages/Uintah/Core/Grid/Patch.h
M Packages/Uintah/Core/Grid/AMR.h
Hack to remove PGI compiler warnings about unreachable code.
M Packages/Uintah/Core/Grid/Grid.h
Use white space... it is also your friend. Seriously (on a related
note ;) can someone tell me why they think "x=3;" is better than "x = 3;"?
M Packages/Uintah/Core/Grid/Variables/ReductionVariable_special.cc
Wow, this is an ugly file... Had to add some (templated) functions
that the PGI compiler was requiring.
M Packages/Uintah/Core/Grid/Variables/ComputeSet.h
Don't seem to need to #include the .cc file (for the templates) with
PGI anymore... at least not on Ranger using pgCC version 7.1-2.
M Packages/Uintah/Core/Grid/Variables/VarTypes.h
Needs the 'long long' var type.
M Packages/Uintah/Core/Grid/Variables/ComputeSet_special.cc
More use of white space, indentation, etc.
M Packages/Uintah/Core/DataArchive/DataArchive.cc
Added {} around what looks like a necessary block of code... strange
that this didn't cause hangs, etc in the code. Perhaps this section
of code isn't used that often? Hope the RT doesn't gripe
tomorrow... it's not my fault!
M Packages/Uintah/Core/Disclosure/TypeUtils.h
M Packages/Uintah/Core/Disclosure/TypeUtils.cc
Added 'long long' function. Organized the code a tad bit to make it
easier to see which functions (for which types) are defined.
M Packages/Uintah/testprograms/sub.mk
M Packages/Uintah/testprograms/PatchBVH/sub.mk
M Packages/Uintah/testprograms/TestFastMatrix/sub.mk
M Packages/Uintah/testprograms/BNRRegridder/sub.mk
M Packages/Uintah/testprograms/IteratorTest/sub.mk
Untabify. Clean up. Removed double definition of LIBS. Requires
BLAS_LIBRARY to link.
M Packages/Uintah/tools/BlasLapackTests/test_blas.c
Ranger's version of testcase... not sure I should have checked it
in... Probably need to have both with the #if separating them.
M Packages/Uintah/Dataflow/Modules/Selectors/ParticleFieldExtractor.h
Fix compiler warning about hidden function (update_progress).
M Packages/Uintah/Dataflow/Modules/Selectors/ParticleFieldExtractor.cc
Fix includes (better ordering). Fix (hack) PGI problem with finding atoll.
M Packages/Uintah/Dataflow/Modules/Selectors/FieldExtractor.h
More white space. Remove unused var. Move common code to above 'if'.
4 lines of code changed in 1 file:
Use the proper c++ standard for include files for the C inlude files, i.e.
<stdio.h> --> <cstdio>
<math.h> --> <cmath>
Tested on g++ 4.2 and 4.3 compilers.
3691 lines of code changed in 3 files:
Fix compiler build problems with gcc-4.3
4 lines of code changed in 2 files:
M CCA/Ports/ProblemSpecInterface.h
M CCA/Components/ProblemSpecification/ProblemSpecReader.h
M CCA/Components/ProblemSpecification/ProblemSpecReader.cc
* Only validate when told too.. currently only sus.cc tells the reader
to validate.
* Default dbg spew to off.
M CCA/Components/Schedulers/SchedulerCommon.cc
Turn off memory use dbg spew.
M StandAlone/sus.cc
Turn on reader validation.
M StandAlone/compare_uda.cc
Purely cosmetic update. Removed tabs, and indented.
M CCA/Components/LoadBalancers/DynamicLoadBalancer.cc
M StandAlone/tools/compare_mms/ExpMMS.h
M StandAlone/tools/compare_mms/SineMMS.h
M StandAlone/tools/compare_mms/MMS.h
M StandAlone/tools/compare_mms/compare_mms.cc
M StandAlone/tools/compare_mms/LinearMMS.h
Fixed compiler warnings.
1 lines of code changed in 2 files:
Major updates:
* Memory checking moved into the Scheduler... currently it is 'turned
off' and thus still only checks memory at the end of a
timestep.. However, we can easily now turn on checking after every
task. Note, this only catches slightly more memory use then
before... what we need to do is have each (some?) task make the memory
tracking call explicitly at a point where the author thinks that the
most memory is being used. The implementation of this isn't yet clear
as the tasks don't have a pointer to the scheduler... Still thinking
on the best way to do this.
* Updated VarTracker code to be a little better with respect to
letting the user know what is going on.
M CCA/Components/SimulationController/SimulationController.h
M CCA/Components/SimulationController/SimulationController.cc
Moved the memory tracking code to the Scheduler.
M CCA/Components/Schedulers/SendState.h
M CCA/Components/Schedulers/SendState.cc
Purely cosmetic (white space, use "{}") changes.
M CCA/Components/Schedulers/OnDemandDataWarehouse.h
setRestarted() is a virtual function.
M CCA/Components/Schedulers/OnDemandDataWarehouse.cc
Purely cosmetic (white space, use "{}") changes. Long lines are ok,
extremely long lines are too much.
M CCA/Components/Schedulers/MPIScheduler.cc
Added /* commented out */ section that will check memory use after each task.
M CCA/Components/Schedulers/SchedulerCommon.h
M CCA/Components/Schedulers/SchedulerCommon.cc
* Moved memory tracking code here.
* Expanded the VarTracker code to be more verbose and to give more
info, check error conditions, etc.
* Indented.
M CCA/Components/Arches/Mixing/NewStaticMixingTable.cc
Only print out the mixing table info on proc 0... makes it much easier to read.
M CCA/Ports/Scheduler.h
Moved memory tracking into Scheduler.
M Core/Grid/Variables/ReductionVariable.h
Cosmetics.
M Core/Grid/Variables/ReductionVariableBase.h
print() is a const function.
M StandAlone/sus.cc
Updates for mem tracking move to scheduler.
6 lines of code changed in 1 file:
Enhanced the load balancer, documentation has been updated to reflect these changes.
Here is a summary of the changes:
Cost weights are now profiled at runtime. This should lead to much more accurate load balances.
The patch assignment algorithm expands upon the orignal greedy algorithm by iterating on the load balance and tightening the constraints at every iteration. The processors each seperatly search a different set of constraints in order to search the space as fast as possible. I believe this produces the optimal load balance.
18 lines of code changed in 2 files: